形变医学图像配准方法设计与仿真

Design and Simulation of Deformation Medical Image Registration Method

利用目前方法对形变医学图像进行配准时,没有提取形变医学图像特征,存在特征点获取结果与实际结果相差大、医学图像配准效果差和医学图像配准时间长的问题。为此提出基于角点检测与SIFT的形变医学图像配准方法。采用角点检测与SIFT相结合的方法对医学图像的特征点进行提取,在图像特征提取前,优先对尺度空间的极值点进行检测,其次生成角点特征,通过检测结果与最终特征点的方向完成医学图像特征点的提取,提升了医学图像配准精度。将提取的特征输入到构建的深度学习模型中,根据提取特征的训练及损失函数的优化实现形变医学图像配准。实验结果表明,通过对上述方法进行特征点获取结果与实际结果对比测试、医学图像配准效果测试和配准时间测试,验证了上述方法的准确性与有效性。

When using the current methods to register the deformable medical images,the deformable medical image features are not extracted,and there are some problems,such as large differences between the feature point acquisition results and the actual results,poor medical image registration effect and long medical image registration time.Therefore,a deformable medical image registration method based on corner detection and SIFT is proposed.The combination of corner detection and SIFT is used to extract the feature points of medical images.Before image feature extraction,the extreme points in scale space are first detected,and then corner features are generated.The feature points of medical images are extracted through the detection results and the direction of the final feature points,which improves the accuracy of medical image registration.The extracted features are input into the constructed deep learning model,and the deformation medical image registration is realized according to the training of the extracted features and the optimization of the loss function.The experimental results show that the accuracy and effectiveness of this method are verified by comparing the feature point acquisition results with the actual results,the medical image registration effect test and the registration time test.